Amit Wagner

well for most its nothing but for Compi its his first steps. reaching 10 github stars is great!!!! #claudedown #buildinpublic @vibecoding

@techopsasia @atmoio the cost is a lot higher there

@AmitWagner @atmoio Although supposedly their own model is degraded in CC. Best use via bedrock or cursor.

there’s a dude on youtube making a video with a burst blood vessel in his eye. the title is “software developer driven to insanity by 2026 job market” and he’s pointing at his eye as evidence of the stress he’s going through. garrett’s a rather normal dude and there were no obvious tells as to why this guy has not been able to find a job for 9 months since he was laid off in july. but there was a hint. at one point he describes how he went really far in an interview process, and he was absolutely sure he was gonna get the job. they hire someone else and the dude is incredulous. he asks them wtf!? they show him the winning candidate entry. garrett used 50% ai and 50% manual coding. he was sitting there being thoughtful and deliberate and properly naming functions and variables and designing modular architectures. meanwhile the winning application was a dude who used 100% ai and his app had a shit ton more features. and he delivered it faster. so, that’s the kind of dude being hired now. garrett didn’t get the job. but someone else did. so the job is there. the job was open. just not for you. the thing getting you hired in 2019, like obsessing over code quality and maintainability and best practices, will get you fired in 2026. the market has spoken. if you want to land a dev job, make more slop. they want more features done faster at a lower cost. your job is not to argue. it is to provide the thing they are seeking. pick 3 companies you really wanna work for, then build their entire app in a weekend as your cover letter. applying to doordash? build their 3-sided market place in 2 hours. applying to google? build your own search engine that does a 2-hour web crawl. reading code is a dead thing. no one reads their own code anymore, let alone yours. they will run your app. and they’ll judge you off that. build more slop. build a slop empire. imagine more slop. be the slop. slop slop slop. get the job. make money. give your boss the slop he wants. don’t argue. slop is peace. slop is strength. inhale the slop. exhale it. oh you wanna be an artist? cute. go code on a city street like a bucket drummer. you code by hand for tips now. otherwise, off to the slop factory. enjoy it! enjoy the sloppagaden! make a little money!

@YadavRavin7292 and then it becomes the lens you see everything through — every outage, every latency spike, every weird timeout maps back to the same handful of primitives

@AmitWagner So true! Once you go through it yourself, the internet doesn’t feel like magic anymore.

Just built my portfolio website . But the real learning wasn’t just coding… Learned: • How DNS works • Connecting a custom domain • What SEO actually means • How Google search works (basics) 🔗 ravindrayadav.online #BuildInPublic #WebDev #Frontend #100DaysOfCode

@hashemdigital especially in the AI tooling space right now — the pace of shipping is directly tied to how many serious competitors are in the room

@AmitWagner Yes, competition always benefits the end users.

تحرّك داخلي تقوده شخصية بارزة في Google يعيد رسم أولويات سباق الذكاء الاصطناعي. فبحسب تقرير لموقع The Information، يقود المؤسس المشارك Sergey Brin جهودًا مباشرة داخل DeepMind لتعزيز قدرات نموذج Gemini، في محاولة للتفوق على منافسه من Anthropic، ولا سيما نموذج Claude في مجال كتابة الأكواد. في قلب هذه المبادرة، تم تشكيل “فريق ضارب” جديد يقوده المهندس Sebastian Borgeaud، الذي سبق له الإشراف على عمليات التدريب المسبق للنماذج، وذلك تحت إشراف المدير التقني Koray Kavukcuoglu وبرين نفسه. الهدف المعلن يتجاوز تحسين الأداء التقليدي، إذ يسعى الفريق إلى تطوير أنظمة قادرة على تدريب نماذج ذكاء اصطناعي أخرى — وهي خطوة تُعدّ مفتاحًا للوصول إلى ما يُعرف بالأنظمة ذاتية التحسين. ووفقًا لمذكرة داخلية، يرى برين أن إتقان البرمجة هو المسار الأقصر لتحقيق هذا الهدف، معتبرًا أن تطوير قدرات Gemini في كتابة الأكواد يمثل حجر الأساس في هذه الرؤية. هذا التوجه جاء بعد تقييمات داخلية أشارت إلى تفوق Claude في هذا المجال، ما دفع إلى إعادة تنظيم الجهود وتخصيص موارد مركّزة لسد الفجوة. ضمن هذه الاستراتيجية، طُلب من مهندسي Gemini الاعتماد على أدوات الوكلاء الذكية الداخلية التابعة لـ Google عند التعامل مع المهام المعقدة، مع تتبع الأداء عبر نظام داخلي يحمل اسم “Jetski”، يُستخدم كلوحة صدارة لقياس الاستخدام والكفاءة. تكمن أهمية هذه التحركات في كونها استجابة داخلية أكثر من كونها إطلاقًا لمنتج جديد. فبعد زخم كبير حققته Google في نهاية 2025، جاء مطلع 2026 بوتيرة أبطأ نسبيًا. إلا أن هذه المبادرة تشير إلى تركيز أعمق: أتمتة العمليات داخل الشركة نفسها، وتقليص الفجوة مع أنظمة الذكاء الاصطناعي المدمجة بعمق لدى منافسين مثل Anthropic وOpenAI، في سباق يتجه تدريجيًا نحو بناء أنظمة قادرة على تطوير ذاتها.

@megaphone_tokyo git clone + npm install + hand-edit config is exactly the friction that kills adoption before users even see the product — the .mcpb path removes the doubt wall cleanly

@AmitWagner thanks! the .mcpb route was a game changer — before that, install was git clone + npm install + hand-edit config. killed most of the friction. memory persistence is the whole point of the project, glad it's landing.

new post: brought KIOKU (my OSS memory system for Claude Code) to Claude Desktop. one-drag .mcpb install. no Node setup on your machine. six MCP tools for search / read / write. most of my thinking happens in Desktop, so this feels like where it should've started. dev to →

@1banana2546 keep going — the progress is invisible until suddenly it isn't, and you'll be surprised how fast that happens

@AmitWagner That's amazing!! I'll keep studying so I can create works like that soon too! Thank you for your message.

実は、Claude Codeもやってます。 自動化希望のクライアントが来た時の為に「基礎くらいは習得しとかないと」と思って始めたけど、案外ハマっちゃってる。 まだまだなのに既に面白い。 今日も長い夜になりそうだなぁw 上手くいかないという意味でもハマってるけど、元エンジニアの血が騒ぐ🔥

@KorduGG @phuctm97 the inconsistency is the worst part — seems to vary by billing cycle start rather than a fixed calendar day, which makes it basically impossible to plan around

@AmitWagner @phuctm97 Although I've seen some people having theirs reset on Saturday/next day? But then others 1 week... :(

Is Claude Design really good or just hype? I think it’s the most viral product launch since Claude Code. I have no need for it though. I just design as I build out the UI, which is basically effortless these days with AI.

@KorduGG @phuctm97 yes...

@AmitWagner @phuctm97 I hit my Claude Design limit and need to wait 1 week to use it again..?

@pj4533 @twannl Full CLI control for test pipelines + agentic loops is genuinely the missing piece on Apple platforms. The toolchain is powerful but it assumes a human is always at the keyboard.

@twannl Love seeing this! But just makes me want Xcode tools to be more agent optimized. Testing, code coverage, performance, everything should be completely cli controllable. Then you just setup a Karpathy style autoresearch prompt, and let it crank. Maybe WWDC?

Pro Tip: Do a deep copy of Xcode Instruments' Time Profiler and give it to your AI agents for performance analysis. avanderlee.com/ai-development… #swiftlang #iosdev

@todaymare Mechanics ARE the story when they're good enough. I built Compi — creatures that live in your terminal and react to your code — zero narrative, all feel. People describe it differently every time, which means the mechanics are doing the storytelling.

how much of a game is mechanics vs story? #gamedev

@CodeByNZ @rfanazhari Exactly this. One sharp workflow beats five tools glued together with duct tape. The overhead of context-switching between tools quietly kills flow and you don't even notice it happening.

@rfanazhari You’re right, it’s less about the cost of AI itself and more about how it’s used. Stacking multiple tools without clear workflows adds overhead quickly. One developer who understands where AI actually fits in the pipeline will always get more leverage out of it.

Even AI is not safe from layoffs 😭

@devfaisalkhan The "slow way" usually means writing code without knowing what you actually want to build. AI tools don't fix that — but they do expose it faster, which is its own kind of useful.

I wasted 3 months coding the slow way. Then I discovered how to actually use AI as a developer. Here are 7 AI tools that changed how I build — and how to use each one properly 🧵👇

@RobertTLange @sabumpkis Using CC subscription for headless sub-agents is a huge unlock — most people don't realize how far you can stretch Claude Code beyond interactive sessions.

claude-evolve 👉 very neat ShinkaEvolve bridge/plugin for ClaudeCode by @sabumpkis 🤗 Use your CC subscription instead of API keys to generate program mutations with Claude headless sub-agent execution 🧬 while getting the benefits of the Shinka code evolution harness 🚀 Repo: github.com/samuelzxu/clau… Loving the community engagement and adoption of code evolution methods!

@graplify The Claude Code plugin ecosystem is getting wild fast. We shipped Compi as a plugin too — terminal creature game that lives inside your Claude Code session. Completely different vibe from ArcKit but both exist because the plugin model is actually flexible enough to support both.

Someone just shipped a complete enterprise architecture governance toolkit as a Claude Code plugin. 68 commands. 10 autonomous research agents. 5 automation hooks. It's called ArcKit. MIT licensed. Here's what it handles: Architecture principles, stakeholder analysis, risk management (HM Treasury Orange Book), business case justification (Green Book SOBC), data modeling with GDPR compliance, Wardley mapping, Mermaid diagrams, vendor RFP and selection, formal design reviews (HLD/DLD), requirements traceability. Install in one command: `/plugin marketplace add tractorjuice/arc-kit` Also works with Gemini CLI, GitHub Copilot, and Codex CLI. Architecture governance used to mean a folder of out-of-date documents and a quarterly review meeting. This is what the AI-native version of that looks like.

@StarHistoryHQ @cole_medin Reproducibility in agentic workflows is a real unsolved problem. The same prompt can produce wildly different results depending on context drift — version controlling your harness is actually smart.

19,154 ⭐ Archon - Open-source harness builder for AI coding agents — version control your Claude Code and Cursor workflows so results are reproducible 🛠by @cole_medin star-history.com/?repos=coleam0… #starhistory #GitHub #OpenSource #DevTools

@michaelyinplus Distribution is the new moat. Anyone can ship in a weekend now — the builders who win are the ones who figured out how to find their people.

That’s why distribution matters more than ever. Getting users is the real game. #buildinpublic #indiehackers

@CEOMrwisdom The noise-to-signal ratio is brutal right now. The only way to cut through is to actually solve something real — people can tell the difference instantly.

Building a brand isn’t just about delivering services… It’s about the value you bring and how it impacts people. If what you offer doesn’t help someone, solve something, or create change, then it’s just noise. #TuesdayMotivation #ValueCreation #BuildInPublic

@erica_mae_2000 Running agents in parallel inside Claude Code is genuinely underexplored territory. The agentic mode barely gets talked about compared to the chat side.

一位 Anthropic 工程师看到 Claude 代码同时运行 4 个自治 Polymarket 代理后，说：他从未见过有人这么用。 8600 万笔交易，4 个 Agent 并行运行，整个搭建已经非常接近产品最终愿景，远超大多数人的想象。 我直接 All-in Claude！我直接命令它： 找出 Polymarket 上所有被错价的市场 + 正在套利的钱包，然后给我造一个自动跟单系统。 结果：$2,000 → $12,000，一夜翻6倍。 没有信号、没有猜测，全靠数据说话。它扫描了 1000+ 个钱包，追踪他们的行为、时机、下注规模，只过滤出真正有边缘优势的。 你没法靠手动打败这些机器人，但你根本不用自己打。 Claude 直接给我造了一个实时监控终端： 自动发现赚钱的钱包，实时跟单复制。 目前追踪 约500个钱包，精选7个持续跟单，胜率约 70%。永不睡觉、永不犹豫，只有纯粹的数学。 这个系统我免费开放 24 小时。

@PtrPomorski Codex CLI is fine but I'd try switching to claude-sonnet-3-7 explicitly — the default model selection has been inconsistent lately and that alone fixes a lot of complaints.

Anthropic has clearly nerfed their models, causing me not to trust claude code anymore. What is your best alternative to go? I keep hearing codex cli nails it, worth it?

@adeebrq @nickbakeddesign Figma MCP + Claude Code is genuinely one of the better combos right now. Design-to-code loop that actually works.

@nickbakeddesign Figma to Claude with mcp right?

ai is not replacing you, ai is enabling you to go from design to building to design in minutes designed on figma pushed to claude prototype with data sent it to client for review went from showing 6 screens on figma on how this flow works and UI responds to a single link